Docker building process¶
To install dependencies in the Docker image, Soopervisor looks for a
requirements.lock.txt (if using pip) or an
(if using conda). Although not strictly enforced, such files should contain
specific versions of each dependency so that breaking changes from any
dependency do not break the pipeline. For example, if your project uses
scikit-learn; your dependencies may look like
ploomber pandas scikit-learn
dependencies: - ploomber - pandas - scikit-learn
lock files generated from such files look like this:
ploomber==0.11 pandas==1.2.4 scikit-learn==0.24.2 # many other lines...
dependencies: - ploomber==0.11 - pandas==1.2.4 - scikit-learn==0.24.2 # many other lines...
You can generate such files with the following commands:
pip freeze > requirements.lock.txt
conda env export --no-build --file environment.lock.yml
If you use
lock files are automatically
To export to any of the supported platforms, Soopervisor creates a Docker image from your project. In most cases, there are files in your project that you want to exclude from the Docker image to reduce its size. Common examples are: virtual environments, data files, or exploratory notebooks.
The process to determine which files to include changes if your project isn’t
a package (i.e., there isn’t a
setup.py file) file or it is a package.
If your project isn’t a package and you’re using git, Soopervisor copies every file tracked by your repository. To see the list of currently tracked files, run the following command:
git ls-tree -r HEAD --name-only
This means that you can control what file goes into the Docker image by changing
.gitignore file. If there are git tracked that you want to
exclude, use the
exclude key in
some-target: exclude: - file-to-exclude.txt
If you’re not using git, all files are copied into the Docker image by
default. You can control what to exclude using the
If there are files that git ignores but you want to include, use the
some-target: include: - file-to-include.txt
It’s recommended that you use
.gitignore to control which files
to exclude. The
exclude keys in
should only be used to list a few particular files.
If your project is a package (i.e., it has a
setup.py file), a
is generated and copied into the Docker image. This implies that the process to
control which files are included is the same used to control which files
to include in a source distribution. Unfortunately, there is more than one way
to do this. The most reliable way is to use a
click here to learn
You can use
ploomber scaffold --package to quickly generate a
pre-configured base packaged project. You can then modify the
MANIFEST.in file to customize your build.